Search papers, labs, and topics across Lattice.
NeuronSoup introduces a novel neural computation architecture that utilizes asynchronous, delay-mediated signal propagation through a pool of shared neurons, moving away from traditional synchronous processing. By employing a genetic algorithm to co-evolve the network's topology, weights, and connectivity, the system achieves 85.9% accuracy on the MNIST digit classification task after 10,000 generations, utilizing a compact model size of 115 KB. This approach not only eliminates the need for a differentiable computation graph but also allows for adaptive computation depth and the discovery of lateral interactions between processing pathways, addressing key limitations of existing deep learning architectures.
Evolving neural architectures can achieve high accuracy without backpropagation, leveraging shared neurons and asynchronous processing to unlock new computational capabilities.
We present NeuronSoup, a neural computation architecture that replaces synchronous layer-by-layer processing with asynchronous, delay-mediated signal propagation through a pool of shared neurons. Each path in the network routes a continuous-valued signal from one input neuron to one output neuron through a variable number of intermediate hidden neurons. Hidden neurons are physically shared across paths: when two paths pass through the same neuron, the second arrival encounters the accumulated state left by the first, producing constructive or destructive interference that depends on signal polarity and arrival timing. The entire architecture -- topology, weights, delays, and connectivity -- is co-evolved by a genetic algorithm operating on a flat real-valued genome of 14,602 genes. On 10-class MNIST digit classification using frozen ResNet18 features as input, the system evolves a network of 204 active paths through 266 hidden neurons (156 shared across multiple paths, with one neuron participating in 11 distinct paths) and achieves 85.9\% test accuracy after 10,000 generations. The trained model occupies 115 KB. We argue that this architecture addresses fundamental limitations of current deep learning: it requires no differentiable computation graph, adapts its computation depth per-sample, and discovers lateral interactions between processing pathways that current architectures must engineer explicitly. We discuss why genetic algorithms are the correct optimization tool for this problem class, why CMA-ES fails at this scale, and how the architecture generalizes to arbitrary domains by substituting the encoder and output structure.